Estimating the hydrogen adsorption in depleted shale gas reservoirs for kerogens in underground hydrogen storage using machine learning algorithms

• Published in: Fuel (Guildford) Vol. 388; p. 134534
• Main Authors: Mwakipunda, Grant Charles, Nafouanti, Mouigni Baraka, Ibrahim, AL-Wesabi, Yu, Long
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 15.05.2025
• Subjects: Hydrogen adsorption; Kerogen; Machine learning; Underground hydrogen storage; Underground hydrogen storage; Kerogen; Machine learning; Hydrogen adsorption
• ISSN: 0016-2361
• DOI: 10.1016/j.fuel.2025.134534

•ABC-MERF algorithm was developed for hydrogen adsorption estimation for UHS purpose.•GWO-MERF was compared with MERF and RF.•ABC-MERF outperformed MERF and RF.•Kerogen KIID types are more favorable for UHS than the other types.•Pressure has been found to have the great impact on hydrogen adsorption. Underground hydrogen storage (UHS) is considered a key technology for the large-scale storage and transport of hydrogen, playing a critical role in the transition to a sustainable energy future. Efficient hydrogen adsorption measurements in geological formations ensure that sufficient hydrogen can be stored safely and economically, making it essential to assess the storage potential of different reservoirs accurately. This study investigated the application of a novel machine learning (ML) model for estimating hydrogen adsorption in shale gas reservoirs, a crucial parameter for assessing their suitability for UHS. The established model, a hybrid of artificial bee colony (ABC) optimization and mixed effects random forest (MERF), denoted as ABC-MERF was utilized. The model performance was evaluated using a dataset obtained from the literature, encompassing key influencing factors such as temperature (T), pressure (P), adsorbed methane (ΓCH4), hydrogen-to-carbon ratio (H/C), oxygen-to-carbon ratio (O/C), and kerogen density (KD). ABC-MERF outperformed ML models such as MERF and random forest (RF) in accuracy, as evidenced by a high coefficient of determination (R2) of 0.9932, a low root mean squared error (RMSE) of 0.0055, and a minimal mean absolute error (MAE) of 0.0007. Furthermore, the study explored the influence of kerogen type on hydrogen adsorption. It was revealed that kerogen type KIID exhibits a greater hydrogen storage propensity than the other kerogen types (KIA, KIIA, KIIIA, KIIB, KIIC). This information provides valuable insights for selecting optimal reservoir formations for UHS projects. Significantly, ABC-MERF substantially reduced computational time compared to other models, requiring only 1.12 s for execution. These findings suggest the ABC-MERF model is a viable and efficient alternative for estimating hydrogen adsorption in shale gas reservoirs. Its high accuracy and reduced computational demand offer substantial advantages in assessing reservoir suitability for UHS, potentially lowering both the time and cost. The investigation of kerogen type further enhances the selection process for suitable UHS reservoirs. This work paves the way for the accelerated development and implementation of UHS solutions, contributing to a more sustainable energy future towards achieving net-zero carbon emissions.